CN118544748A - Adaptive control method of vehicle electronic suspension based on preceding vehicle intention recognition - Google Patents

Abstract

The present invention relates to an adaptive control method for an electronically controlled suspension of a vehicle based on the recognition of the intention of a preceding vehicle, comprising the following steps: acquiring multi-source data, and acquiring state data of a vehicle and a preceding vehicle based on the multi-source data; predicting the brake pedal displacement and speed of the preceding vehicle based on response surface modeling; proposing an optimized Transformer model to identify the driving intention of the vehicle; establishing a dynamic model to calculate the pitch angular velocity of the vehicle body; solving the expected active force of the electronically controlled suspension through the pitch angular velocity deviation; establishing a shock absorber damping model based on an oil-gas flow equation; and determining an optimal control current according to a current-external characteristic curve. The present invention can simultaneously consider the influence of the driving intentions of the vehicle and the preceding vehicle on the pitch vibration of the vehicle body, and also simultaneously take factors such as safety, comfort, smoothness and maneuverability into consideration, so that the established suspension control model is more intelligent and has a certain reference value for the formulation of a vehicle body pitch suppression strategy.

Description

Adaptive control method of vehicle electronic suspension based on preceding vehicle intention recognition

Technical Field

The invention relates to the technical field of vehicle electronically controlled suspension, and in particular to an adaptive control method of a vehicle electronically controlled suspension based on preceding vehicle intention recognition.

Background Art

In recent years, with the development of smart car technology, cars can be controlled by automatic driving systems in the longitudinal direction. Consumers pay more and more attention to longitudinal comfort performance, and their requirements are getting higher and higher. The "head-up" and "head-nodding" phenomenon of the vehicle during braking/acceleration, and the resulting pitch angle changes are one of the key factors affecting ride comfort. Especially in urban road conditions, frequent vehicle starts and stops can easily cause significant body pitch vibrations, affecting passenger ride comfort and even causing adverse reactions such as motion sickness.

In addition to deceleration and impact, the cause of discomfort during braking/acceleration is the large pitch motion of the vehicle. The suspension is closely related to the pitch motion of the vehicle. Since the 20th century, various types of springs, dampers and suspensions with certain flexibility in all directions have been developed. The simplest and most common type of suspension is passive suspension, but it cannot directly control the pitch of the vehicle. With the development of modern control theory and electrical technology, the electronically controlled suspension can actively adjust the damping force according to the working conditions, generate corresponding force or movement, offset unnecessary body movements, and effectively improve the smoothness, safety and handling stability of the car.

The existing technology integrates open-loop optimization control and robust feedback control to form a suspension control strategy. It uses the current vehicle status information monitored by on-board sensors to calculate the desired control force in real time through the controller and adjust the working point of the damper, thereby suppressing the pitch motion of the vehicle body under acceleration and deceleration conditions.

In the suspension system, the calculation response of the controller and actuator has a time delay. Using the current vehicle motion state, the system output force will be delayed, and the control of the vehicle body pitch motion is limited. In the following condition, the motion state of the leading vehicle has a great impact on the vehicle. The existing technology ignores the effect of the leading vehicle, resulting in a violent pitch motion of the vehicle when the leading vehicle stops suddenly.

Summary of the invention

The main technical problem to be solved by the present invention is to provide an adaptive control method for vehicle electronic suspension based on the recognition of the intention of the preceding vehicle, so as to realize the pre-control of the pitch motion of the vehicle body under the acceleration and deceleration conditions of the vehicle, so as to solve the shortcomings of the traditional electronic suspension system in terms of control lag, and provide a reference for formulating a more reasonable and accurate vehicle body pitch control strategy, and also provide technical support for the optimization of gear shifting smoothness and safety.

To achieve the above object, the present invention provides the following technical solutions:

A vehicle electronically controlled suspension adaptive control method based on preceding vehicle intention recognition, the method comprising the following steps:

S1, obtaining multi-source data from vehicle sensors, vehicle cameras and road-side 3D cameras, obtaining state data of the vehicle and the preceding vehicle based on the multi-source data, fitting the state data according to a simplified UniTire model, and obtaining a road adhesion coefficient μ, wherein the state data includes vehicle speed, acceleration, road characteristics, slip rate k, sideslip angle α and tire force;

S2, predicting the brake pedal displacement and speed of the preceding vehicle based on response surface modeling;

S3, propose an optimized Transformer model to identify vehicle driving intention;

S4, establishing a dynamic model to calculate the vehicle body pitch angular velocity;

S5, solving the expected active force of the electronically controlled suspension through the pitch angular velocity deviation;

S6. Establish a shock absorber damping model based on the oil and gas flow equation;

S7. Determine the optimal control current based on the current-external characteristic curve.

As a further technical solution of the present invention, in step S1, the step of obtaining multi-source data from the vehicle-mounted sensor, the vehicle-mounted camera and the road-side three-dimensional camera, obtaining the state data of the vehicle and the preceding vehicle based on the multi-source data, fitting the state data according to the simplified UniTire model, and obtaining the road adhesion coefficient μ comprises:

Multi-source data is obtained from vehicle-mounted sensors, vehicle-mounted cameras and road-side three-dimensional cameras, and data with a forward sensing range of 500m and a time window of 10 minutes are selected for preprocessing, and the preprocessing includes single-source data filtering and fusion; in order to reduce the influence of noise and make the signal smooth, the single-source data is first filtered by a narrow stopband filter to eliminate the influence of factors such as device interface and transmission line; then a fourth-order Butterworth bandpass filter is used to eliminate the influence of noise, and a moving average method of a 30ms sliding window is used for smoothing; multi-sensor fusion data can provide more comprehensive information for vehicle intention recognition, but the redundancy of multi-source data will reduce the calculation efficiency; in order to balance data accuracy and calculation efficiency, the present invention uses kernel principal component analysis (KPCA) and cumulative contribution rate _rccr to determine the data dimension, and obtains 12-dimensional state data with time series, including the speed, acceleration, road characteristics and tire state data of the vehicle and the front vehicle, and the tire state data includes slip rate k, side slip angle α and tire force, wherein the tire force includes Fx, Fy and Fz;

The simplified UniTire model is used to fit the tire status data over a period of time to obtain the adhesion coefficient μ. The simplified UniTire model is shown below:

Among them, F _nx means the ratio of the longitudinal force F _x to the tire load F _z . The longitudinal force F _nx is a function of the road adhesion coefficient μ and the longitudinal slip rate S _x . The partial derivative expression of F _nx with respect to S _x is as follows:

Formula 2:

In formula 2, f(S _x ,μ,K _xn , _Ex ) is expressed as:

Formula 3:

By simulating the simplified UniTire model, the parameters K _xn = 20, _Ex = 0.05 are determined, and the tire state parameters are nonlinearly fitted using the gradient descent method to obtain the road adhesion coefficient μ.

As a further technical solution of the present invention, in step S2, the step of predicting the brake pedal displacement and speed of the preceding vehicle based on response surface modeling includes:

The response surface model is obtained through real vehicle tests: the target vehicle that meets the test conditions is determined, and the target vehicle is controlled to travel according to the preset control strategy under multiple preset conditions to obtain the data of the generalized displacement of the pedal and the braking force/driving force changing with the acceleration at multiple vehicle speeds; when establishing the response surface model, in order to obtain a reliable and high-precision response surface model and avoid high-order vibration, the SequentialReplacement method with a lower cost is selected to explore the order of the response surface model. After analysis, the second-order response surface model is selected, and the model is as follows:

Formula 4:

Among them, _ps is the acceleration/brake pedal displacement, _β0 , _β1 , _β2 , _β3 are the coefficients to be calibrated, R is the residual term to be calibrated, _x1 , _x2 , _x3 are the vehicle speed, acceleration, and road adhesion coefficient respectively; the experimental results are used as training data to obtain the model coefficients and response surface model; the error calculation method is used to analyze the model accuracy, and the root mean square error is 1.6e-16, indicating that the constructed response surface approximate model meets the accuracy requirements. Under a certain vehicle speed, the corresponding acceleration/brake pedal displacement can be accurately calculated by the deceleration; the differential method is used to calculate the pedal displacement speed based on the pedal displacement, and the calculation formula is as follows:

Formula 5:

Wherein, p _v is the pedal displacement speed, and t is the time; the vehicle speed, acceleration, and pedal displacement and speed are obtained through steps S1 and S2, providing basic data for step S3 to identify the driver's desired intention.

As a further technical solution of the present invention, in step S3, the step of proposing an optimized Transformer model to identify vehicle driving intention includes:

In order to detect all hidden relationships between features in multi-source datasets, an enhanced local attention mechanism ELAM is proposed, which matches the most relevant features and relationships in local semantics through two causal convolution windows; input n-step e-dimensional time series features value, value, The value is calculated as follows:

Formula 6:

Formula 7:

Formula 8:

in, and is the kernel size of the random convolution operation [1, k _s ], W ^q , W ^v , is the learning parameter matrix; the softmax function is used to standardize the weights, and the ELAM output is:

Formula 9:

Formula 10: O _e =Concat(Att ₁ ,…,Att _i ,…,Att _hn )W ^o ;

Produced by causal convolution value, After the value is calculated, the ELAM and transformer models are integrated to optimize the local attention. The optimized transformer model is:

Formula 11:

in, is the predicted driving force/braking force, The decoding layer consists of n and a decoder. The structure of VT decoder is adopted for each decoding layer. The multi-source fusion data set obtained in step S1 is divided into training set and validation set according to 7:3. A comparative experiment is designed to determine the length of the historical domain and the prediction domain. The range of both is set to 100ms-1000ms. The prediction accuracy is evaluated by rmse and ^R2 . The calculation formula is as follows:

Formula 12:

Formula 13:

Among them, _yi and are the actual value and predicted value of driving force/braking force, and l _s are the average value and prediction domain length; through analysis, the history domain length has little effect on the model accuracy, so 500ms is selected according to computer performance; when the prediction domain length is 200ms, the proposed optimized transformer model can achieve accurate prediction of brake pressure with an accuracy of more than 90%.

As a further technical solution of the present invention, in step S4, the step of establishing a dynamic model to calculate the vehicle body pitch angular velocity includes:

A five-degree-of-freedom vehicle vertical dynamics model considering pitch motion is established. The five degrees of freedom are the vertical degrees of freedom of the front and rear suspensions, the vertical degrees of freedom of the front and rear wheels, and the pitch rotation degrees of freedom of the suspension. The dynamic model is as follows:

Formula 14:

in, is the vertical acceleration of the vehicle body center of mass, is the vehicle body pitch angular acceleration, and is the front and rear vehicle body vertical acceleration, and is the vertical acceleration of the front and rear unsprung masses, F _mf and F _mr are the equivalent forces generated by the braking intensity, and the calculation formula is:

Formula 15:

Where z represents the braking strength, _hg is the height of the center of mass of the vehicle body, and L is the wheelbase. In the five-degree-of-freedom half-car model, _Ff and _Fr are the expected control forces output by the front and rear suspensions, and the calculation formula is:

Formula 16:

in, They are the front unsprung mass, the front vehicle body, the rear unsprung mass, and the vertical velocity of the rear vehicle body. The vertical motion of the vehicle causes the tire load to change. By adding the tire load change to the half-vehicle model, the actual tire load calculation formula is:

Where, G = ( _mb + _mwr + _mmf ) g, g is the acceleration due to gravity; when the vehicle moves at a variable speed in a straight line, the longitudinal motion of the vehicle satisfies Newton's second law;

Formula 18: F _b +F _f +F _w +F _i =δma _m ;

Among them, _Fb is the vehicle braking/driving force, _Ff is the friction resistance, _Fw is the wind resistance, _Fi is the slope resistance, δ is the vehicle rotation mass conversion coefficient, and the calculation formula is:

Formula 19:

Among them, i _g is the transmission ratio of the transmission, i ₀ is the main reducer ratio, η _T is the transmission efficiency, I _w is the wheel moment of inertia, and _If is the flywheel moment of inertia; the braking/driving strength is an important parameter that links the longitudinal motion and vertical motion of the vehicle, and the calculation formula is:

Formula 20: z = _-am /g;

Based on the five-degree-of-freedom half-car model, the pitch angular velocity of the car body during the speed change process is calculated through the braking/driving intensity and the expected acceleration, providing a data basis for the subsequent establishment of a control model.

As a further technical solution of the present invention, in step S5, the step of solving the expected active force of the electronically controlled suspension by the pitch angular velocity deviation includes:

In order to make the vehicle body pitch motion smoother, the nominal value of the pitch angular velocity is taken as Then the pitch angular velocity input deviation of the controller is:

Formula 21:

Using the geometric relationship of sprung mass motion, the pitch angular velocity deviation is solved into the velocity deviation of the sprung mass at the front and rear axles, that is:

Formula 22:

Therefore, the expression of the desired force of the active suspension force actuator at the front and rear axles based on the Skyhook controller is:

Formula 23:

Discretize it and let the control period be ΔT, then:

Formula 24:

To further calculate the skyhook control law, take the front wheel as an example, substitute Equation 22 into Equation 23, and arrange it as follows:

Formula 25:

The skyhook control law is derived using the pitch angular velocity, which mainly includes three parts, namely the control of the pitch angular displacement, the control of the pitch angular velocity and the control of the pitch angular acceleration, ultimately achieving the suppression of the vehicle body pitch motion.

As a further technical solution of the present invention, in step S6, the step of establishing a shock absorber damping model based on the oil-gas flow equation includes:

The modeling process of the shock absorber is mainly based on the relevant knowledge of fluid mechanics, and the modeling process of the compression stroke and the recovery stroke is similar, so only the recovery stroke is used as an example for explanation. The damping force calculation formula of the CDC shock absorber is expressed as:

Formula 26: F _f =A _p -A _r )P ₁ -A _p P ₂ ;

Where A _p and A _r are the effective areas of the piston and piston rod, respectively, P ₁ and P ₂ are the pressures of the recovery chamber and compression chamber; under the action of the exciting speed v _f , the changes of the outflow flow Q1 of the recovery chamber and the inflow flow Q2 of the compression chamber in the shock absorber are expressed as follows:

Formula 27:

The structure and modeling process of the restoring valve and the compensation valve are similar to those of the conventional shock absorber. They are divided into two working states: before and after the valve is opened. The corresponding flow pressure difference relationship is:

Formula 28:

Among them, _Cd is the valve port flow coefficient, _dfg and _dbg are the fixed throttle diameters of the restoring valve and the compensating valve, _dfv and _dbv are the outer radii of the restoring valve disc and the compensating valve disc, _ωf and _ωb are the deformations of the restoring valve disc and the compensating valve disc after the valve is opened. After the oil flows through the one-way valve disc, it flows into the compensation chamber through the main valve and the pilot valve, so:

Formula 29:

Among them, d _cv is the outer radius of the one-way valve disc, z is the opening of the one-way valve disc, and the flow equations of the damping hole H, the main valve and the pilot valve are respectively;

Formula 30:

Among them, _dmv and _dpv are the effective diameters of the damping holes H in front of the main valve and the pilot valve respectively, y and x are the displacements of the main valve core and the pilot valve core respectively. By combining the above equations, the oil pressure of each chamber can be solved, and the relationship between the restoring stroke damping force and the exciting velocity _vf (or displacement S) can be obtained by substituting them into the equations.

As a further technical solution of the present invention, in step S7, the step of determining the optimal control current according to the current-external characteristic curve includes:

The damping regulating valve of the electronically controlled shock absorber usually adopts a two-stage or three-stage proportional electromagnetically driven overflow valve. As the core component of the shock absorber, the electromagnetic force characteristics of the electromagnetically driven overflow valve have a significant impact on the dynamic response of the shock absorber. The characteristic model is shown below:

Formula 31:

in, is the Hamiltonian operator, J is the current density, μ is the magnetic permeability of the medium, and _Aθ is the vector magnetic potential of the solenoid valve in the θ coordinate direction; the vector magnetic potential A is obtained by calculation, and according to its relationship with the magnetic induction intensity:

Formula 32:

Formula 33:

Among them, _Br and _Bz are the magnetic induction intensities of the solenoid valve in the r and z coordinate directions, respectively. When solving the model, A is obtained by numerical interpolation, and the magnetic induction intensity B at the air gap is further obtained, and finally the calculation formula of the electromagnetic force is obtained:

Formula 34:

Among them, _μ0 is the vacuum magnetic permeability, S is the cross-sectional area of the magnetic circuit, a is the correction coefficient with an empirical value of 3 to 5, and δ is the working air gap length; a current-external characteristic test is carried out based on the characteristic model, and a certain control current of the shock absorber is input within the rated working current range (0 to 1.7A) of the electronically controlled shock absorber. The damping force under different excitation speeds is tested to obtain the relationship among the control current, damping force and excitation speed; based on the current-external characteristic curve, the control current value corresponding to the expected optimal damping is obtained.

Compared with the prior art, the present invention has the following beneficial effects:

The proposed adaptive control method of vehicle electronic suspension based on leading vehicle intention recognition can simultaneously consider the impact of the driving intentions of the vehicle and the leading vehicle on the pitch vibration of the vehicle body, while taking factors such as safety, comfort, smoothness and maneuverability into account. The established suspension control model is more intelligent and has certain reference value for the formulation of vehicle body pitch suppression strategy.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a flow chart of a method for adaptively controlling a vehicle electronically controlled suspension based on preceding vehicle intention recognition according to an embodiment of the present invention.

FIG. 2 is a schematic diagram of a five-degree-of-freedom half-car model provided by an embodiment of the present invention.

DETAILED DESCRIPTION

In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

Referring to FIG. 1 and FIG. 2 , as an embodiment of the present invention, the present invention proposes a vehicle electronically controlled suspension adaptive control method based on preceding vehicle intention recognition, the method comprising the following steps:

S1, obtaining multi-source data from vehicle sensors, vehicle cameras and road-side 3D cameras, obtaining state data of the vehicle and the preceding vehicle based on the multi-source data, fitting the state data according to a simplified UniTire model, and obtaining a road adhesion coefficient μ, wherein the state data includes vehicle speed, acceleration, road characteristics, slip rate k, sideslip angle α and tire force;

Specifically, multi-source data is obtained from vehicle sensors, vehicle cameras and road-side three-dimensional cameras, and data with a forward sensing range of 500m and a time window of 10 minutes are selected for preprocessing, and the preprocessing includes single-source data filtering and fusion; in order to reduce the influence of noise and make the signal smooth, the single-source data is first filtered by a narrow stopband filter to eliminate the influence of factors such as device interface and transmission line; then a fourth-order Butterworth bandpass filter is used to eliminate the influence of noise, and a moving average method of a 30ms sliding window is used for smoothing; multi-sensor fusion data can provide more comprehensive information for vehicle intention recognition, but the redundancy of multi-source data will reduce the calculation efficiency; in order to balance data accuracy and calculation efficiency, the present invention adopts kernel principal component analysis (KPCA) and cumulative contribution rate _rccr to determine the data dimension, and obtains 12-dimensional state data with time series, including the speed, acceleration, road characteristics and tire state data of the vehicle and the front vehicle, and the tire state data includes slip rate k, side slip angle α and tire force, wherein the tire force includes Fx, Fy and Fz;

The simplified UniTire model is used to fit the tire status data over a period of time to obtain the adhesion coefficient μ. The simplified UniTire model is shown below:

Formula 1:

Among them, F _nx means the ratio of the longitudinal force F _x to the tire load F _z . The longitudinal force F _nx is a function of the road adhesion coefficient μ and the longitudinal slip rate S _x . The partial derivative expression of F _nx with respect to S _x is as follows:

Formula 2:

In formula 2, f(S _x ,μ,K _xn , _Ex ) is expressed as:

Formula 3:

By simulating the simplified UniTire model, the parameters K _xn = 20, _Ex = 0.05 are determined, and the tire state parameters are nonlinearly fitted using the gradient descent method to obtain the road adhesion coefficient μ.

S2, predicting the brake pedal displacement and speed of the preceding vehicle based on response surface modeling;

Based on the obtained data such as vehicle speed, acceleration and road adhesion coefficient μ, the pedal displacement response surface model is used to obtain the generalized displacement and speed of the pedal, providing input data for the recognition of the speed change intention of the leading vehicle and the self-vehicle.

The response surface model is obtained through real vehicle tests: the target vehicle that meets the test conditions is determined, and the target vehicle is controlled to travel according to the preset control strategy under multiple preset conditions to obtain the data of the generalized displacement of the pedal and the braking force/driving force changing with the acceleration at multiple vehicle speeds; when establishing the response surface model, in order to obtain a reliable and high-precision response surface model and avoid high-order vibration, the SequentialReplacement method with a lower cost is selected to explore the order of the response surface model. After analysis, the second-order response surface model is selected, and the model is as follows:

Formula 4:

Among them, _ps is the acceleration/brake pedal displacement, _β0 , _β1 , _β2 , _β3 are the coefficients to be calibrated, R is the residual term to be calibrated, _x1 , _x2 , _x3 are the vehicle speed, acceleration, and road adhesion coefficient respectively; the experimental results are used as training data to obtain the model coefficients and response surface model; the error calculation method is used to analyze the model accuracy, and the root mean square error is 1.6e-16, indicating that the constructed response surface approximate model meets the accuracy requirements. Under a certain vehicle speed, the corresponding acceleration/brake pedal displacement can be accurately calculated by the deceleration; the differential method is used to calculate the pedal displacement speed based on the pedal displacement, and the calculation formula is as follows:

Formula 5:

Wherein, p _v is the pedal displacement speed, and t is the time; the vehicle speed, acceleration, and pedal displacement and speed are obtained through steps S1 and S2, providing basic data for step S3 to identify the driver's desired intention.

S3, propose an optimized Transformer model to identify vehicle driving intention;

In order to detect all hidden relationships between features in multi-source datasets, an enhanced local attention mechanism ELAM is proposed, which matches the most relevant features and relationships in local semantics through two causal convolution windows; input n-step e-dimensional time series features value, value, The value is calculated as follows:

Formula 6:

Formula 7:

Formula 8:

in, and is the kernel size of the random convolution operation [1, k _s ], W ^q , W ^v , is the learning parameter matrix; the softmax function is used to standardize the weights, and the ELAM output is:

Formula 9:

Formula 10: O _e =Concat(Att ₁ ,…,ATT _i ,…,Att _hn )W ^o ;

Produced by causal convolution value, After the value is calculated, the ELAM and transformer models are integrated to optimize the local attention. The optimized transformer model is:

Formula 11:

in, is the predicted driving force/braking force, The decoding layer consists of n and a decoder. The structure of VT decoder is adopted for each decoding layer. The multi-source fusion data set obtained in step S1 is divided into training set and validation set according to 7:3. A comparative experiment is designed to determine the length of the historical domain and the prediction domain. The range of both is set to 100ms-1000ms. The prediction accuracy is evaluated by rmse and ^R2 . The calculation formula is as follows:

Formula 12:

Formula 13:

Among them, _yi and are the actual value and predicted value of driving force/braking force, and l _s are the average value and prediction domain length; through analysis, the history domain length has little effect on the model accuracy, so 500ms is selected according to computer performance; when the prediction domain length is 200ms, the proposed optimized transformer model can achieve accurate prediction of brake pressure with an accuracy of more than 90%.

S4, establishing a dynamic model to calculate the vehicle body pitch angular velocity;

A five-degree-of-freedom vehicle vertical dynamics model considering pitch motion is established. The five degrees of freedom are the vertical degrees of freedom of the front and rear suspensions, the vertical degrees of freedom of the front and rear wheels, and the pitch rotation degrees of freedom of the suspension. The dynamic model is as follows:

Formula 14:

in, is the vertical acceleration of the vehicle body center of mass, is the vehicle body pitch angular acceleration, and is the front and rear vehicle body vertical acceleration, and is the vertical acceleration of the front and rear unsprung masses, F _mf and F _mr are the equivalent forces generated by the braking intensity, and the calculation formula is:

Formula 15:

Where z represents the braking strength, _hg is the height of the center of mass of the vehicle body, and L is the wheelbase. In the five-degree-of-freedom half-car model, _Ff and _Fr are the expected control forces output by the front and rear suspensions, and the calculation formula is:

Formula 16:

in, They are the front unsprung mass, the front vehicle body, the rear unsprung mass, and the vertical velocity of the rear vehicle body. The vertical motion of the vehicle causes the tire load to change. By adding the tire load change to the half-vehicle model, the actual tire load calculation formula is:

Where G = ( _mb + _mwr + _mmf ) g, g is the acceleration of gravity; when the vehicle moves at a variable speed in a straight line, the longitudinal motion of the vehicle satisfies Newton's second law;

Formula 18: F _b +F _f +F _w +F _i =δma _m ;

Among them, _Fb is the vehicle braking/driving force, _Ff is the friction resistance, _Fw is the wind resistance, _Fi is the slope resistance, δ is the vehicle rotation mass conversion coefficient, and the calculation formula is:

Formula 19:

Among them, i _g is the transmission ratio of the transmission, i ₀ is the main reducer ratio, η _T is the transmission efficiency, I _w is the wheel moment of inertia, and _If is the flywheel moment of inertia; the braking/driving strength is an important parameter that links the longitudinal motion and vertical motion of the vehicle, and the calculation formula is:

Formula 20: z = _-am /g;

Based on the five-degree-of-freedom half-car model, the pitch angular velocity of the car body during the speed change process is calculated through the braking/driving intensity and the expected acceleration, providing a data basis for the subsequent establishment of a control model.

S5, solving the expected active force of the electronically controlled suspension through the pitch angular velocity deviation;

In order to make the vehicle body pitch motion smoother, the nominal value of the pitch angular velocity is taken as Then the pitch angular velocity input deviation of the controller is:

Formula 21:

Using the geometric relationship of sprung mass motion, the pitch angular velocity deviation is solved into the velocity deviation of the sprung mass at the front and rear axles, that is:

Formula 22:

Therefore, the expression of the desired force of the active suspension force actuator at the front and rear axles based on the Skyhook controller is:

Formula 23:

Discretize it and let the control period be ΔT, then:

Formula 24:

To further calculate the skyhook control law, take the front wheel as an example, substitute Equation 22 into Equation 23, and arrange it as follows:

Formula 25:

The skyhook control law is derived using the pitch angular velocity, which mainly includes three parts, namely the control of the pitch angular displacement, the control of the pitch angular velocity and the control of the pitch angular acceleration, ultimately achieving the suppression of the vehicle body pitch motion.

S6. Establish a shock absorber damping model based on the oil and gas flow equation;

The modeling process of the shock absorber is mainly based on the relevant knowledge of fluid mechanics, and the modeling process of the compression stroke and the recovery stroke is similar, so only the recovery stroke is used as an example for explanation. The damping force calculation formula of the CDC shock absorber is expressed as:

Formula 26: F _f =(A _p -A _r )P ₁ -A _p P ₂ ;

Where A _p and A _r are the effective areas of the piston and piston rod, respectively, P ₁ and P ₂ are the pressures of the recovery chamber and compression chamber; under the action of the exciting speed v _f , the changes of the outflow flow Q1 of the recovery chamber and the inflow flow Q2 of the compression chamber in the shock absorber are expressed as follows:

Formula 27:

The structure and modeling process of the restoring valve and the compensation valve are similar to those of the conventional shock absorber. They are divided into two working states: before and after the valve is opened. The corresponding flow pressure difference relationship is:

Formula 28:

Among them, _Cd is the valve port flow coefficient, _dfg and _dbg are the fixed throttle diameters of the restoring valve and the compensating valve, _dfv and _dbv are the outer radii of the restoring valve disc and the compensating valve disc, _ωf and _ωb are the deformations of the restoring valve disc and the compensating valve disc after the valve is opened. After the oil flows through the one-way valve disc, it flows into the compensation chamber through the main valve and the pilot valve, so:

Formula 29:

Among them, d _cv is the outer radius of the one-way valve disc, z is the opening of the one-way valve disc, and the flow equations of the damping hole H, the main valve and the pilot valve are respectively;

Formula 30:

Among them, _dmv and _dpv are the effective diameters of the damping holes H in front of the main valve and the pilot valve respectively, y and x are the displacements of the main valve core and the pilot valve core respectively. By combining the above equations, the oil pressure of each chamber can be solved, and the relationship between the restoring stroke damping force and the exciting velocity _vf (or displacement S) can be obtained by substituting them into the equations.

S7. Determine the optimal control current based on the current-external characteristic curve.

The damping regulating valve of the electronically controlled shock absorber usually adopts a two-stage or three-stage proportional electromagnetically driven overflow valve. As the core component of the shock absorber, the electromagnetic force characteristics of the electromagnetically driven overflow valve have a significant impact on the dynamic response of the shock absorber. The characteristic model is shown below:

Formula 31:

in, is the Hamiltonian operator, J is the current density, μ is the magnetic permeability of the medium, and _Aθ is the vector magnetic potential of the solenoid valve in the θ coordinate direction; the vector magnetic potential A is obtained by calculation, and according to its relationship with the magnetic induction intensity:

Formula 32:

Formula 33:

Among them, _Br and _Bz are the magnetic induction intensities of the solenoid valve in the r and z coordinate directions, respectively. When solving the model, A is obtained by numerical interpolation, and the magnetic induction intensity B at the air gap is further obtained, and finally the calculation formula of the electromagnetic force is obtained:

Formula 34:

Among them, _μ0 is the vacuum magnetic permeability, S is the cross-sectional area of the magnetic circuit, a is the correction coefficient with an empirical value of 3 to 5, and δ is the working air gap length; a current-external characteristic test is carried out based on the characteristic model, and a certain control current of the shock absorber is input within the rated working current range (0 to 1.7A) of the electronically controlled shock absorber. The damping force under different excitation speeds is tested to obtain the relationship among the control current, damping force and excitation speed; based on the current-external characteristic curve, the control current value corresponding to the expected optimal damping is obtained.

It should be understood that, for those skilled in the art, the principles of the present invention and the above description can be improved or transformed, or the method provided by the present invention can be applied to similar aerial image recognition tasks, and all these improvements and transformations should fall within the scope of protection of the claims attached to the present invention.

Claims (8)

1. A vehicle electronically controlled suspension adaptive control method based on preceding vehicle intention recognition, characterized in that the method comprises the following steps: S1, obtaining multi-source data from vehicle sensors, vehicle cameras and road-side 3D cameras, obtaining state data of the vehicle and the preceding vehicle based on the multi-source data, fitting the state data according to a simplified UniTire model, and obtaining a road adhesion coefficient μ, wherein the state data includes vehicle speed, acceleration, road characteristics, slip rate k, sideslip angle α and tire force; S2, predicting the brake pedal displacement and speed of the preceding vehicle based on response surface modeling; S3, propose an optimized Transformer model to identify vehicle driving intention; S4, establishing a dynamic model to calculate the vehicle body pitch angular velocity; S5, solving the expected active force of the electronically controlled suspension through the pitch angular velocity deviation; S6. Establish a shock absorber damping model based on the oil and gas flow equation; S7. Determine the optimal control current based on the current-external characteristic curve. 2. The vehicle electronic suspension adaptive control method based on the preceding vehicle intention recognition according to claim 1 is characterized in that, in step S1, the multi-source data is obtained by the vehicle-mounted sensor, the vehicle-mounted camera and the road-side three-dimensional camera, the state data of the vehicle and the preceding vehicle are obtained based on the multi-source data, and the state data are fitted according to the simplified UniTire model to obtain the road adhesion coefficient μ, which comprises: Multi-source data is obtained from vehicle sensors, vehicle cameras and road-side 3D cameras. Data with a forward sensing range of 500m and a time window of 10 minutes are selected for preprocessing, which includes single-source data filtering and fusion. First, the single-source data is filtered using a narrow stopband filter, and then a fourth-order Butterworth bandpass filter is used to eliminate the influence of noise, and a moving average method with a 30ms sliding window is used for smoothing. The kernel principal component analysis method and the cumulative contribution rate _{rccr are} used to determine the data dimension, and 12-dimensional state data with time series are obtained. The state data includes the speed, acceleration, road characteristics and tire state data of the vehicle and the front vehicle. The tire state data includes slip rate k, sideslip angle α and tire force, wherein the tire force includes Fx, Fy and Fz. The simplified UniTire model is used to fit the tire state data to obtain the adhesion coefficient μ. The simplified UniTire model is shown below: Formula 1: Among them, F _nx means the ratio of the longitudinal force F _x to the tire load F _z . The longitudinal force F _nx is a function of the road adhesion coefficient μ and the longitudinal slip rate S _x . The partial derivative expression of F _nx with respect to S _x is as follows: Formula 2: In formula 2, f(S _x ,μ,K _xn , _Ex ) is expressed as: Formula 3: By simulating the simplified UniTire model, the parameters K _xn = 20, _Ex = 0.05 are determined, and the tire state parameters are nonlinearly fitted using the gradient descent method to obtain the road adhesion coefficient μ. 3. The vehicle electronic suspension adaptive control method based on preceding vehicle intention recognition according to claim 1, characterized in that in step S2, the step of predicting the brake pedal displacement and speed of the preceding vehicle based on response surface modeling comprises: The response surface model is obtained through real vehicle tests: the target vehicle that meets the test conditions is determined, and the target vehicle is controlled to travel according to the preset control strategy under multiple preset conditions, and the data of the generalized displacement of the pedal and the braking force/driving force changing with acceleration at multiple vehicle speeds are obtained; when establishing the response surface model, the Sequential Replacement method is selected to explore the selection of the second-order response surface model, and the model is shown below: Formula 4: Among them, _ps is the acceleration/brake pedal displacement, _β0 , _β1 , _β2 , _β3 are the coefficients to be calibrated, R is the residual term to be calibrated, _x1 , _x2 , _x3 are the vehicle speed, acceleration, and road adhesion coefficient respectively; the experimental results are used as training data to obtain the model coefficients and response surface model; the error calculation method is used to analyze the model accuracy, and the corresponding acceleration/brake pedal displacement is accurately calculated by the deceleration under the determined vehicle speed; the differential method is used to calculate the pedal displacement speed according to the pedal displacement, and the calculation formula is as follows: Formula 5: Where p _v is the pedal displacement speed and t is the time. 4. The vehicle electronic suspension adaptive control method based on preceding vehicle intention recognition according to claim 1 is characterized in that, in step S3, the step of proposing an optimized Transformer model to recognize the vehicle driving intention comprises: An enhanced local attention mechanism ELAM is proposed, which matches the most relevant features and the relations in local semantics through two causal convolution windows; inputs n-step e-dimensional time series features value, value, The value is calculated as follows: Formula 6: Formula 7: Formula 8: in, and is the kernel size of the random convolution operation [1, k _s ], W ^q , W ^v , is the learning parameter matrix; the softmax function is used to standardize the weights, and the ELAM output is: Formula 9: Formula 10: O _e =Concat(Att ₁ ,…,Att _i ,…,Att _hn )W ^o ; Produced by causal convolution value, After the value is calculated, the ELAM and transformer models are integrated to optimize the local attention. The optimized transformer model is: Formula 11: in, is the predicted driving force/braking force, The decoding layer consists of n and a decoder. The structure of VT decoder is adopted for each decoding layer. The multi-source fusion data set obtained in step S1 is divided into training set and validation set according to 7:3. A comparative experiment is designed to determine the length of the historical domain and the prediction domain. The range of both is set to 100ms-1000ms. The prediction accuracy is evaluated by rmse and ^R2 . The calculation formula is as follows: Formula 12: Formula 13: Among them, _yi and are the actual value and predicted value of driving force/braking force, and l _s are the mean and predicted domain lengths. 5. The vehicle electronic suspension adaptive control method based on preceding vehicle intention recognition according to claim 1, characterized in that in step S4, the step of establishing a dynamic model to calculate the vehicle body pitch angular velocity comprises: A five-degree-of-freedom vehicle vertical dynamics model considering pitch motion is established. The five degrees of freedom are the vertical degrees of freedom of the front and rear suspensions, the vertical degrees of freedom of the front and rear wheels, and the pitch rotation degrees of freedom of the suspension. The dynamic model is as follows: Formula 14: in, is the vertical acceleration of the vehicle body center of mass, is the vehicle body pitch angular acceleration, and is the front and rear vehicle body vertical acceleration, and is the vertical acceleration of the front and rear unsprung masses, F _mf and F _mr are the equivalent forces generated by the braking intensity, and the calculation formula is: Formula 15: Where z represents the braking strength, _hg is the height of the center of mass of the vehicle body, and L is the wheelbase. In the five-degree-of-freedom half-car model, _Ff and _Fr are the expected control forces output by the front and rear suspensions, and the calculation formula is: Formula 16: in, They are the front unsprung mass, the front vehicle body, the rear unsprung mass, and the vertical velocity of the rear vehicle body. The vertical motion of the vehicle causes the tire load to change. By adding the tire load change to the five-degree-of-freedom half-vehicle model, the actual tire load calculation formula is: Where, G = ( _mb + _mwr + _mmf ) g, g is the acceleration due to gravity; when the vehicle moves at a variable speed in a straight line, the longitudinal motion of the vehicle satisfies Newton's second law; Formula 18: F _b +F _f +F _w +F _i =δma _m ; Among them, _Fb is the vehicle braking/driving force, _Ff is the friction resistance, _Fw is the wind resistance, _Fi is the slope resistance, δ is the vehicle rotation mass conversion coefficient, and the calculation formula is: Formula 19: Among them, i _g is the transmission ratio of the transmission, i ₀ is the main reducer ratio, η _T is the transmission efficiency, I _w is the wheel moment of inertia, and _If is the flywheel moment of inertia; the braking/driving strength is an important parameter that links the longitudinal motion and vertical motion of the vehicle, and the calculation formula is: Formula 20: z = _-am /g; Based on the five-degree-of-freedom half-car model, the pitch angular velocity of the car body during the speed change process is calculated through the braking/driving intensity and the expected acceleration. 6. The vehicle electronically controlled suspension adaptive control method based on preceding vehicle intention recognition according to claim 5, characterized in that, in step S5, the step of solving the expected active force of the electronically controlled suspension by using the pitch angular velocity deviation comprises: In order to make the vehicle body pitch motion smoother, the nominal value of the pitch angular velocity is taken as Then the pitch angular velocity input deviation of the controller is: Formula 21: Using the geometric relationship of sprung mass motion, the pitch angular velocity deviation is solved into the velocity deviation of the sprung mass at the front and rear axles, that is: Formula 22: Therefore, the expression of the desired force of the active suspension force actuator at the front and rear axles based on the skyhook controller is: Formula 23: Discretize it and let the control period be ΔT, then: Formula 24: To further calculate the ceiling control law, take the front wheel as an example, substitute Equation 22 into Equation 23, and rearrange it to obtain: Formula 25: The ceiling control law is derived using the pitch angular velocity, which mainly includes three parts, namely the control of the pitch angular displacement, the control of the pitch angular velocity and the control of the pitch angular acceleration, ultimately achieving the suppression of the pitch motion of the vehicle body. 7. The vehicle electronic suspension adaptive control method based on preceding vehicle intention recognition according to claim 1 is characterized in that, in step S6, the step of establishing a shock absorber damping model based on the oil-gas flow equation comprises: The damping force calculation formula of CDC shock absorber is expressed as: Formula 26: F _f =(A _p -A _r )P ₁ -A _p P ₂ ; Wherein, _Ap and _Ar are the effective areas of the piston and piston rod respectively, _P1 and _P2 are the pressures of the recovery chamber and compression chamber respectively; under the action of the exciting velocity _vf , the changes of the outflow flow Q1 of the recovery chamber and the inflow flow Q2 of the compression chamber in the shock absorber are respectively expressed as: Formula 27: The structure and modeling process of the restoring valve and the compensation valve are similar to those of the conventional shock absorber. They are divided into two working states: before and after the valve is opened. The corresponding flow pressure difference relationship is: Formula 28: Among them, _Cd is the valve port flow coefficient, _dfg and _dbg are the fixed throttle diameters of the restoring valve and the compensating valve, _dfv and _dbv are the outer radii of the restoring valve disc and the compensating valve disc, _ωf and _ωb are the deformations of the restoring valve disc and the compensating valve disc after the valve is opened. After the oil flows through the one-way valve disc, it flows into the compensation chamber through the main valve and the pilot valve, so: Formula 29: Among them, d _cv is the outer radius of the one-way valve disc, z is the opening of the one-way valve disc, and the flow equations of the damping hole H, the main valve and the pilot valve are respectively; Formula 30: Wherein, _dmv and _dpv are the effective diameters of the damping holes H in front of the main valve and the pilot valve, respectively, and y and x are the displacements of the main valve core and the pilot valve core, respectively. 8. The vehicle electronically controlled suspension adaptive control method based on preceding vehicle intention recognition according to claim 1, characterized in that in step S7, the step of determining the optimal control current according to the current-external characteristic curve comprises: The electromagnetic force characteristics of the electromagnetically driven overflow valve have a significant impact on the dynamic response of the shock absorber. The characteristic model is shown below: Formula 31: in, is the Hamiltonian operator, J is the current density, μ is the magnetic permeability of the medium, and _Aθ is the vector magnetic potential of the solenoid valve in the θ coordinate direction; the vector magnetic potential A is obtained by calculation, and according to its relationship with the magnetic induction intensity: Formula 32: Formula 33: Among them, _Br and _Bz are the magnetic induction intensities of the solenoid valve in the r and z coordinate directions, respectively. When solving the model, A is obtained by numerical interpolation, and the magnetic induction intensity B at the air gap is further obtained, and finally the calculation formula of the electromagnetic force is obtained: Formula 34: Among them, _μ0 is the vacuum magnetic permeability, S is the cross-sectional area of the magnetic circuit, a is the correction coefficient, and δ is the working air gap length; a current-external characteristic test is carried out based on the characteristic model. Within the rated working current range of the electronically controlled shock absorber, a certain control current of the shock absorber is input, and the damping force under different excitation speeds is tested to obtain the relationship among the control current, damping force and excitation speed; according to the current-external characteristic curve, the control current value corresponding to the expected optimal damping is obtained.